The main difference is that I have spent a lot of time working on writing code quickly and correctly. Every time I competed in TC I would read the code of the best people and try to figure out why their code was so much simpler than mine. Now I think I can write clean code faster than most.
The downside is that I have spent a ton of time working on 50-200 line projects. Throw me into a 500k line code base and I'm no longer world class.
Getting to where I am is mostly a product of spending a ton of time in college practicing for programming contests. In college I saw a very strong correlation between time spent programming outside of class and programming skill.
1) Talented: Can they code well, even when things get tricky? This loosely correlates with things like TopCoder rank, but it leans more towards software engineering and debugging skills.
2) Driven: Are they relentlessly productive, even without supervision?
3) Solves the right problem: If there's a solution which provides 80% of the benefits for 20% of the cost, do they find it? Do they also know when it's time to turn a nasty ball of hacks into a rigorous model?
If you have these 3 things, then you will easily create 5 times more value than the average programmer. Note that you won't necessary get paid 5 times more, because that also depends on your negotiating skills, your marketing ability, and how much money your clients or employers have.
About your 3 points: I agree that having these in one person is a good combination. However people are rarely exceptional in all 3 points. Some guys are exceptional algorithmizers, others have really good intuition for enginering, creating an amazing design (for code), yet others have exceptional memory so that they can dive into huge bloated shitty code, and yet others deal well with monotony (when you have to do unchallanging tasks, or deal with other people's shitty code). Some people are fast like hell, others can grow beautiful thoughts if you give them days to think about a problem. But I agree that a bit of all of this is needed mostly. Also being strong in one territory you can emulate other strengths. If the tasks are not on your limit, you can emulate fastness with smartness, and emulate smartness with being fast. But when the problems are hard enough that you are at your 'limit', this emulation no longer works.
If you want to earn more than average, here are some suggestions:
1) Contribute heavily to an open source project, preferably one used by people with lots of money. This demonstrates your talent and drive more clearly than a few interview questions.
2) Learn how to market your skills and how to negotiate. For the former, study patio11's career. For the latter, try reading "Getting to Yes" from the Harvard Negotiation Project.
Put them in a foreign environment, with a foreign language/api/framework/application/etc and within a short learning curve they will find solutions to problems 1) you didn't know you had, or 2) you had long since given up on finding a better solution to.
If you can replace the programmer with books (and time to read them), they can't be that valuable. You don't pay one person 5 times market rate if 5 people at market rate can do their job. You pay someone for the concentration of talent.
Also, much of a programmer's worth pays off over time: specifically how easy is it to maintain their code? Writing truly maintainable code is HARD and difficult to quantify.
Yes, and I think you cannot convince them on a standard job interview. A job interview is like this: you solved the problems we gave you, congratulations, we hired you (for slightly more than market rate). They will not give you harder and harder inteview questions and finally give you 5x market salary in my experience.
Earning 5x market rate is not only about capabilities but more about building a very interesting track record, personal brand, which helps you differentiate yourself and helps you convince clients about your value. It is a bit like a long-term strategy game.
Edit: I am really curious who are those extremely smart programmers.
Guys on the top of this list?
http://community.topcoder.com/tc?module=AlgoRank
Or guys who were at the top of the Netflix prize top scores list?
Those are some truly gifted guys (one in a million caliber), but even most of them don't earn 5x market rate.